VIEW THIS AS

Auto mode follows the Route Engine until you choose a viewpoint.

YOU ARE HERE

ROUTE CHECK

CONNECTED TO

WHAT NEXT

Use the canonical route for this room, or HELP if you are unsure.

Ultimate GCE A-Level H2 Mathematics Syllabus Guide | SEAB 9758 Topics, Worked Examples and Study Planning

Three learners review open books together at a classroom table, with stacks of textbooks, stationery and a whiteboard in the bright room.

GCE A-Level H2 Mathematics, syllabus 9758, helps Singapore students turn algebra, graphs, geometry and data into clear mathematical decisions. This guide explains the 2027 course through original worked examples, practical study routes and questions families can use when choosing subjects or planning revision. Start by checking the student’s examination year and subject code, then connect each school topic to the reasoning it develops.

A confident H2 Mathematics journey depends on more than completing worksheets. Students need to recognise a structure, choose a method, carry out accurate working and explain what the answer means. The chapters below help parents understand those demands without having to relearn the whole course, while giving students concrete ways to repair gaps and practise independently. The examination details were checked on 17 November 2026; school teaching sequences and individual subject arrangements should be confirmed with the school.

  • Course: Singapore–Cambridge GCE A-Level H2 Mathematics, SEAB 9758, for the 2027 examination.
  • Main learning routes: functions, sequences, vectors, complex numbers, calculus, probability and statistics.
  • Best starting point: check algebra, functions and basic calculus before increasing question difficulty.
  • Useful evidence of progress: an independently completed solution with a sensible check and an explanation of the chosen method.
  • For parents: discuss subject fit, school feedback and realistic practice time; avoid using one test result as a permanent label.
  • For students: use the worked examples as models, then solve a changed version without looking.

Chapter index

Chapters 1–6
  1. Choose the course by the work it asks you to do
  2. Understand the examination before designing revision
  3. Repair algebra where later topics depend on it
  4. Read functions as rules with an allowed input set
  5. Use sequences to describe repeated change
  6. Make three-dimensional geometry visible
Chapters 7–12
  1. Treat complex numbers as a precise extension of familiar algebra
  2. Differentiate to describe change and make decisions
  3. Integrate to accumulate quantities and recover relationships
  4. Connect approximation with differential-equation models
  5. Build probability from clearly defined events
  6. Choose a discrete model before calculating
Chapters 13–16
  1. Distinguish an individual measurement from a sample mean
  2. Use hypothesis tests to evaluate evidence carefully
  3. Use regression for a defined prediction task
  4. Build a study plan that produces independent solutions

CHAPTER 1 OF 16

Choose the course by the work it asks you to do

Back to contents

H2 Mathematics rewards a particular combination of patience and flexibility. A student might begin with a diagram, translate it into equations, solve those equations and then reject an answer because it makes no sense in the original situation. Another question might provide a collection of measurements and ask whether a claim is supported. Both demand mathematical judgement, although the calculations look very different.

For a family considering H1 or H2, begin with three questions. Does the student enjoy sustained mathematical thinking? How secure are the foundations needed for the available course? What mathematics is required by the specific university programmes the student may later consider? The third question needs checking against the relevant institutions’ current requirements. A broad statement such as “science needs H2” is too imprecise to make an individual decision.

The 2027 H1 Mathematics syllabus is a separate course, not a promise that every H2 topic appears in shorter form. H1 has one three-hour, 100-mark paper, with 40 marks for pure mathematics and 60 for probability and statistics. Its pure mathematics coverage is narrower. Students should compare actual content, school eligibility and intended pathways rather than treating the subject labels as measures of intelligence.

What readiness looks like

A useful readiness conversation concerns observable work. Can the student rearrange an equation without losing restrictions? Can they distinguish a function from its derivative? When a solution fails, can they identify the first doubtful step? These skills are more informative than a confident statement that mathematics is a favourite subject. Enjoyment matters, but it needs to be paired with habits that survive unfamiliar questions.

For a shorter orientation before beginning the teaching chapters, read How to Make Sense of the A-Level H2 Mathematics Syllabus. The present guide extends that overview through calculations, interpretation and practice variations. Students who already understand the course structure can move directly to their weakest mathematical topic.

Suppose a student obtained strong secondary-school marks by recognising familiar question patterns. That success is real. It also leaves a useful next step: practise choosing between methods when the question does not announce its topic. A student with less impressive marks may already explain ideas well but need greater fluency. The two students need different preparation, even if they enter the same class.

Parents can help by asking the school about expectations, support arrangements and any subject-change procedures early. Schools may organise topics differently, so a chapter taught in one college’s first term may appear later elsewhere. Use the school’s actual schedule to plan practice. Treat this guide as a learning map, not a national timetable or an admissions decision.

Contents · Next chapter

CHAPTER 2 OF 16

Understand the examination before designing revision

Back to contents

The compact table below records the central paper arrangements in the 2027 H2 Mathematics syllabus. Read the official document for complete instructions and exclusions.

PaperDurationMarksWeightCoverage
Paper 13 hours10050%Pure mathematics; answer all questions.
Paper 23 hours10050%Section A: pure mathematics, 40 marks. Section B: probability and statistics, 60 marks. Answer all questions.
H2 Mathematics 9758 examination structure

An approved graphing calculator without a computer algebra system is expected. Follow each question’s working requirements; graph-based solutions need sketches. A formula list is provided. The linked MF27 formula list is the version linked by SEAB for these mathematics syllabuses.

Translate the structure into practice

Across the two papers, the allocation is 140 marks of pure mathematics and 60 marks of probability and statistics. This arithmetic helps explain why a student cannot build a balanced revision programme around their favourite statistics chapter alone. It does not prescribe an exact study-time ratio. Someone with severe gaps in algebra may need more foundation work; someone who has neglected statistical interpretation may need a concentrated repair period there.

Three hours for 100 marks gives an average of 1.8 minutes per mark. Use that only as a planning reference. A short interpretation may take less time; an unfamiliar setup may require careful reading before any calculation begins. During practice, record where time goes. “I ran out of time” becomes actionable when it is replaced by “I spent twelve minutes trying to force an unsuitable substitution.”

The calculator should be familiar enough that its operation does not consume the attention needed for reasoning. Practise entering brackets correctly, selecting sensible graph windows, storing unrounded values and identifying the correct distribution parameters. A machine can calculate the probability of the event entered. It cannot know whether that event represents the wording of the question.

When reviewing a timed paper, separate three outcomes: mathematics that was unknown, mathematics that was known but selected too slowly, and mathematics that was selected correctly but executed inaccurately. Each outcome suggests a different next task. Re-reading notes may help the first; mixed questions may help the second; deliberate checking routines may help the third.

For a more focused discussion of the papers and examination resources, use the existing 2027 H2 Mathematics paper, MF27 and graphing-calculator guide. Keep that reference beside the official instructions, while using the worked chapters here to develop the mathematical decisions those resources support.

Contents · Previous chapter · Next chapter

CHAPTER 3 OF 16

Repair algebra where later topics depend on it

Back to contents

Many H2 difficulties begin several lines before the final error. An incorrect differentiation result may come from a mishandled index. A faulty probability calculation may come from reading “at least” as “more than”. A wrong vector answer may begin with a displacement written in the opposite direction. Foundations are useful because they reduce the number of things a student must consciously manage at once.

Worked example: cancellation has conditions

Consider (x2 − 9)/(x − 3). Factorising gives (x − 3)(x + 3)/(x − 3), so the expression equals x + 3 when x ≠ 3. The original expression is undefined at x = 3. Cancelling a factor simplifies the expression where the division is allowed; it does not fill the hole in its original domain.

This small distinction later matters in functions, limits of models and equation solving. If a student reports x + 3 without noticing the restriction, ask them to substitute x = 3 into the original expression. The purpose is to make the missing condition visible, not to demand a memorised warning after every line of working.

Worked example: an inequality is sensitive to sign

Solve (x − 1)/(x + 2) > 0. The critical values are −2 and 1. Test one point in each interval: x = −3 gives a positive ratio; x = 0 gives a negative ratio; x = 2 gives a positive ratio. Therefore x < −2 or x > 1. Exclude −2 because the denominator vanishes, and exclude 1 because the requested inequality is strict.

Multiplying immediately by x + 2 without considering its sign can reverse or preserve the inequality unpredictably. The interval method keeps the reasoning visible. Later, when a graph is used instead, the same restrictions still apply. Different tools should lead to the same mathematical story.

A short foundation check

Try four tasks without a calculator: solve 2x+1 = 16; simplify ln(e3t); differentiate (3x + 1)4; and solve |x − 4| < 2. The answers are x = 3, 3t, 12(3x + 1)3, and 2 < x < 6. Explain why each operation is valid. If an answer is missing, repair that specific idea before turning the check into a long worksheet.

For revision, keep a small page of recurring algebra errors with a correct example beside each one. Include the trigger: negative exponent, nested brackets, logarithm of a product, denominator restriction or inequality sign. A list saying only “careless mistake” is too vague to guide improvement. The repair should identify the move that needs changing.

Contents · Previous chapter · Next chapter

CHAPTER 4 OF 16

Read functions as rules with an allowed input set

Back to contents

A function is more than its formula. The rule tells us how to produce an output; the domain tells us which inputs we are considering. This distinction becomes essential when reversing a function, combining functions or interpreting a model. A height model may be algebraically defined for negative times, while the physical experiment begins at time zero.

Worked example: an inverse needs a unique return journey

Let f(x) = (x − 2)2 + 1 for x ≥ 2. Its smallest output is 1, so its range is y ≥ 1. To find the inverse, write y − 1 = (x − 2)2. Since the original domain gives x − 2 ≥ 0, take the nonnegative square root: x = 2 + √(y − 1). Hence f−1(x) = 2 + √(x − 1), with x ≥ 1.

If the original domain had included every real number, inputs equally far on either side of 2 would produce the same output. For example, f(1) and f(3) would both equal 2. A return rule could not decide which input to choose. The domain restriction is therefore part of the reasoning, not a decorative line attached to the answer.

Composite functions: check the handover

Suppose g(x) = √x for x ≥ 0 and h(x) = x + 1 for all real x. The composite h(g(x)) is defined for every allowed input of g because its outputs can enter h. In the opposite order, g(h(x)) is not a valid composite over the whole stated domain of h: h produces negative outputs when x < −1, which g cannot accept. Decide existence using the stated functions and domains rather than silently changing them.

Graphs should explain the algebra

For y = 1/(x − 2) + 3, the vertical asymptote is x = 2 and the horizontal asymptote is y = 3. The intercepts provide additional anchors: y = 5/2 when x = 0, and x = 5/3 when y = 0. A sketch built from these features is more reliable than one copied loosely from a calculator screen.

When transforming a graph, track a known point. If (a, b) lies on y = f(x), then (a + 2, b − 1) lies on y = f(x − 2) − 1. This coordinate check protects against the common mistake of moving a graph in the wrong horizontal direction. Practise with one point, then a turning point, then an asymptote.

Parents can ask the student to explain the graph in ordinary language: what inputs are allowed, where outputs become large, whether the function is increasing and what a particular point means. A clear explanation usually reveals whether the student has connected the equation and picture.

Contents · Previous chapter · Next chapter

CHAPTER 5 OF 16

Use sequences to describe repeated change

Back to contents

A sequence records values in an order. A series adds values from a sequence. Keeping these jobs separate prevents a surprising number of mistakes. The amount present after the tenth step is not automatically the total amount added during ten steps. Before choosing a formula, write the first few values and say what each one represents.

Worked example: a repeated transfer

Imagine a container begins with 100 units of a substance. At each step, 20% of the current amount is removed and then 30 units are added. If un is the amount after n steps and u0 = 100, then un+1 = 0.8un + 30. The first values are 100, 110, 118 and 124.4.

A possible steady value L would satisfy L = 0.8L + 30, giving L = 150. Subtracting 150 from the recurrence gives un+1 − 150 = 0.8(un − 150). The distance from 150 is therefore multiplied by 0.8 each time, so un = 150 − 50(0.8)n. Since (0.8)n tends to zero, the amounts approach 150.

The calculation is useful because it explains the behaviour. Merely solving for L does not prove every recurrence approaches that value. The multiplier tells us whether departures shrink, grow or alternate. Always connect the limiting value to the actual rule.

Worked example: adding a geometric series

Suppose a display uses 12 lights on its first strip, then each subsequent strip uses half as many in an idealised mathematical model. The first four terms are 12, 6, 3 and 1.5, with sum 22.5. The first n terms sum to 12(1 − 0.5n)/(1 − 0.5). The infinite mathematical series has sum 24 because |0.5| < 1.

Fractional lights show the limit of this physical description. The calculation is a model of diminishing quantities, not a construction instruction. Students should be comfortable saying when the mathematics is exact within a model and when the model stops describing the real object.

A useful practice variation changes the timing: what if the 30 units are added before 20% is removed? The rule becomes un+1 = 0.8(un + 30), or 0.8un + 24. Its steady value is 120. The wording changed only slightly, but the model changed substantially. Reading order is mathematical work.

Contents · Previous chapter · Next chapter

CHAPTER 6 OF 16

Make three-dimensional geometry visible

Back to contents

Vectors allow a geometric problem to be handled through directions, lengths and coordinates. The first decision is what each vector means. A position vector points from the origin to a point. A displacement vector takes us from one point to another. A direction vector describes orientation and need not end at a particular point.

Worked example: direction and distance

Let A = (1, 2, −1) and B = (3, −1, 3). The displacement from A to B is B − A = (2, −3, 4). Its magnitude is √(4 + 9 + 16) = √29. A unit vector in that direction is (2, −3, 4)/√29. Reversing the journey reverses the vector but leaves the distance unchanged.

The line through A in this direction can be written r = (1, 2, −1) + t(2, −3, 4). Setting t = 0 gives A; setting t = 1 gives B. These checks are quick and valuable. They catch a direction vector accidentally used as a starting point or a subtraction made in the wrong order.

Worked example: meet a plane

Take the line r = (1, 0, 2) + t(1, 2, −1) and the plane x + y + z = 6. Substitution gives (1 + t) + 2t + (2 − t) = 6, so 3 + 2t = 6 and t = 1.5. The intersection is (2.5, 3, 0.5). Its coordinates add to 6, which checks the plane condition.

The plane’s normal vector is n = (1, 1, 1). It points perpendicular to the plane. A line’s direction vector points along the line. Confusing these two roles can produce a numerically neat angle with the wrong geometric meaning. When finding a line–plane angle, first identify whether the calculation gives the angle with the normal and whether a complementary angle is needed.

Projection gives a useful explanation

For vectors a = (3, 4, 0) and unit vector n = (1, 0, 0), the scalar product a·n equals 3. This is the signed component of a along the x-direction. The remaining component is (0, 4, 0). Thinking in components makes perpendicular distances and feet of perpendiculars less mysterious.

Before solving a longer vector question, make a rough sketch and label the points, line directions and plane normals. The sketch need not be perspective-perfect. Its job is to keep relationships distinct. Then write one equation for each relationship: lying on a line, lying in a plane, being perpendicular or sharing a point. Geometry becomes manageable when each sentence receives its own mathematical statement.

Contents · Previous chapter · Next chapter

CHAPTER 7 OF 16

Treat complex numbers as a precise extension of familiar algebra

Back to contents

Complex numbers allow us to solve equations that have no real roots. The symbol i satisfies i2 = −1. A complex number a + bi has a real part a and an imaginary part b. Its location on an Argand diagram is the point (a, b). The word “imaginary” does not make the arithmetic arbitrary; it introduces a carefully defined new direction.

Worked example: divide using the conjugate

To simplify (3 + 2i)/(1 − i), multiply numerator and denominator by 1 + i. The numerator becomes 3 + 3i + 2i + 2i2 = 1 + 5i. The denominator becomes 1 − i2 = 2. The result is 1/2 + (5/2)i.

Multiplying back by 1 − i checks the answer: (1/2 + 5i/2)(1 − i) = 3 + 2i. This check also illustrates why a conjugate is useful. The product of a + bi and a − bi is a2 + b2, which is real.

Worked example: roots carry geometric information

For z2 − 4z + 13 = 0, completing the square gives (z − 2)2 = −9. Therefore z = 2 ± 3i. The roots lie at (2, 3) and (2, −3), reflected across the real axis. Their sum is 4 and their product is 13, matching the quadratic’s coefficients.

The modulus of 2 + 3i is √13, its distance from the origin. Its argument is the angle from the positive real axis, with the quadrant handled carefully. A calculator’s inverse tangent output needs interpretation; the signs of the real and imaginary parts determine which region contains the point.

Multiplication by i also has a clean geometric meaning: i(a + bi) = −b + ai. The point (a, b) becomes (−b, a), a quarter-turn anticlockwise about the origin. Test it on 1, which becomes i, before applying it to less familiar values.

Keep this chapter’s practice focused on the course you are taking. Some older resources introduce additional representations and techniques. The 9758 document should decide what belongs in examination preparation. Extra mathematics can be interesting, but it should not displace the core skill of moving accurately between algebra and the Argand diagram.

Contents · Previous chapter · Next chapter

CHAPTER 8 OF 16

Differentiate to describe change and make decisions

Back to contents

A derivative describes how one quantity changes with another. In a graph it gives a tangent gradient. In a model it can describe speed, sensitivity or the direction in which an output changes. The formula is only useful after the variables and their units have been identified.

Worked example: an open box

A square sheet measures 20 cm by 20 cm. Squares of side x cm are cut from its corners, and the sides are folded up. The resulting box has volume V = x(20 − 2x)2, with 0 < x < 10. Expanding gives V = 400x − 80x2 + 4x3.

Differentiate: V′ = 400 − 160x + 12x2 = 4(3x − 10)(x − 10). The stationary values occur at x = 10/3 and x = 10. Only 10/3 is inside the physical domain. Since V′ changes from positive to negative there, the volume is maximised at x = 10/3 cm. The maximum is 16,000/27 cm3, approximately 592.6 cm3.

The domain did real work: x = 10 would flatten the base to zero width. A solution that reports both stationary values as possible box designs has completed the differentiation but not the problem. Always return from the equation to the object.

Implicit relationships

For x2 + y2 = 25, differentiating with respect to x gives 2x + 2y(dy/dx) = 0. Hence dy/dx = −x/y where y ≠ 0. At (3, 4), the tangent gradient is −3/4, so the tangent is y − 4 = −(3/4)(x − 3). Substitution of the point checks the line, while the radius gradient 4/3 provides a perpendicularity check.

Connected rates require a chain of meaning

If a circle’s radius increases at 0.2 cm per second, A = πr2 gives dA/dt = 2πr(dr/dt). When r = 5 cm, the area increases at 2π cm2 per second. The area rate depends on the current radius even though the radius rate is constant. This is why identifying what stays fixed matters.

When a student struggles with calculus applications, ask them to pause before differentiating. What quantity is to be maximised or measured? What variable controls it? What constraints connect the variables? Writing those answers often resolves the problem more effectively than practising another differentiation rule in isolation.

Contents · Previous chapter · Next chapter

CHAPTER 9 OF 16

Integrate to accumulate quantities and recover relationships

Back to contents

Integration connects local change with total accumulation. It also reverses differentiation, although a constant is needed when recovering an indefinite family of functions. A student should know which of these jobs the question is asking them to do. An antiderivative, a signed integral, an area and a volume are related objects with different interpretations.

Worked example: an antiderivative with a visible check

Consider ∫2x/(x2 + 4) dx. The numerator is the derivative of the denominator, so the result is ln(x2 + 4) + C. Differentiating this expression returns 2x/(x2 + 4). Here x2 + 4 is always positive, so the logarithm is defined for every real x.

Recognising the inner function is more dependable than searching for a visual resemblance to a memorised line. If the numerator were x instead of 2x, a factor of 1/2 would be needed. This small variation is a useful test of whether the method has been understood.

Worked example: signed integral versus area

For y = x − 1 between x = 0 and x = 3, the definite integral is [x2/2 − x]03 = 1.5. The total geometric area is different because part of the graph lies below the x-axis. The triangle from 0 to 1 has area 0.5, and the triangle from 1 to 3 has area 2. The total area is 2.5 square units.

A sketch immediately explains the difference: the integral subtracts the below-axis contribution, while total area counts both pieces positively. When a region crosses an axis or two curves exchange positions, split the calculation at the relevant crossing.

Integration by parts as a deliberate exchange

To integrate xex, choose u = x and dv = exdx. Then du = dx and v = ex, giving ∫xexdx = xex − ex + C. The exchange is useful because differentiating x makes the remaining integral simpler. A poor choice can make the problem more complicated, so explain the choice before carrying it out.

For volume of revolution, imagine a thin slice. Rotating y = x about the x-axis from x = 0 to x = 2 produces circular slices of area πx2. Their accumulated volume is π∫02x2dx = 8π/3. The squared radius and units of volume should both appear in the explanation. A formula remembered without its geometry is easy to misapply.

Contents · Previous chapter · Next chapter

CHAPTER 10 OF 16

Connect approximation with differential-equation models

Back to contents

Some functions are awkward to calculate or manipulate directly near a point. A polynomial approximation can reveal their local behaviour. A differential equation works in the other direction: it begins with a rule for change and asks what overall behaviour follows. Both topics reward attention to conditions.

Worked example: a local approximation

Near x = 0, e2x has expansion 1 + 2x + 2x2 + (4/3)x3 + higher-order terms. The coefficients come from substituting 2x into the exponential series. At x = 0.05, the cubic approximation is 1 + 0.1 + 0.005 + 0.0001667 = 1.1051667, close to e0.1 ≈ 1.1051709.

The small difference is informative. It shows what truncation does, but does not establish that every polynomial approximation works equally well at every input. Record the expansion point, the terms retained and any validity conditions. An approximation is a controlled simplification, not an exact identity with the final terms erased.

For (1 + 2x)−1, the geometric expansion is 1 − 2x + 4x2 − 8x3 + …, valid when |2x| < 1. At x = 0.1 the cubic approximation is 0.832, while the exact value is 1/1.2 ≈ 0.833333. The restriction belongs to the input after substitution, so it becomes |x| < 1/2.

Worked example: change proportional to what remains

Let Q(t) be an amount that decreases at a rate proportional to its current value. The model is dQ/dt = −kQ, where k > 0. Separating variables gives dQ/Q = −k dt. Integrating and using Q(0) = 80 gives Q = 80e−kt.

If Q falls to 40 after three hours, then e−3k = 1/2, so k = (ln 2)/3 per hour. After nine hours, Q = 80(1/2)3 = 10. The repeated-halving interpretation checks the exponential calculation.

The negative sign expresses decrease. The initial condition determines the particular member of the solution family. The unit of k ensures that the exponent is dimensionless. These details are not extras; together they explain why the model means what it claims.

Ask whether the model remains sensible

A model that predicts unlimited growth may describe an early phase while ignoring later resource limits. A cooling model might assume a constant surrounding temperature. Students do not need to invent scientific facts beyond a question’s information. They do need to identify the assumptions they used and avoid extending a formula beyond the situation it was built to describe.

Contents · Previous chapter · Next chapter

CHAPTER 11 OF 16

Build probability from clearly defined events

Back to contents

Probability becomes easier when the event is written before the formula. “At least two”, “exactly two” and “two given that one condition holds” are different requests. A tree, table or small sample-space list can expose those differences more clearly than a page of symbols.

Worked example: count the intended selection

A committee of three is chosen uniformly from five students, including A and B. There are 5 choose 3 = 10 possible committees. To include both A and B, choose the third member from the other three students, giving three committees. The probability is therefore 3/10.

Now ask for exactly one of A and B. Choose which of them is included in two ways, then choose two of the other three students in three ways. There are six suitable committees, giving probability 6/10. The remaining committee includes neither A nor B. The counts 3 + 6 + 1 = 10 check the partition of possibilities.

Worked example: conditioning changes the reference group

In an invented group of 100 students, 60 use a bus, 40 use a train, and 20 use both during a typical week. If one bus user is selected at random, the probability that they also use a train is 20/60 = 1/3. If one student is selected from the whole group, the probability of using both is 20/100 = 1/5.

The numerator counts the same 20 students, but the denominator changes because the information changes. This is the heart of conditional probability. A reliable solution explicitly states which group has become the sample space.

Independence is different from mutual exclusion

Two events are mutually exclusive when they cannot occur together. Two events are independent when learning that one occurred does not alter the probability of the other. For events with positive probabilities, mutual exclusion actually prevents independence: once one occurs, the other becomes impossible.

Use simple examples before memorising rules. On one fair die roll, “even” and “odd” are mutually exclusive. On two independent fair die rolls, “first roll is even” and “second roll is even” are independent. The number of experiments and the wording of the events matter.

When checking an answer, ask whether it lies between zero and one, whether a complement would be simpler and whether the result changes sensibly when the conditions change. A probability of 0.9 for an event described as rare should prompt a re-reading of the model, even if every calculator entry was technically correct.

Contents · Previous chapter · Next chapter

CHAPTER 12 OF 16

Choose a discrete model before calculating

Back to contents

A discrete random variable assigns a numerical value to each outcome. The variable is not the same thing as the event that produced it. If X counts successful trials, define what a success means, how many trials occur and why the probability structure is suitable.

Worked example: a binomial count

Suppose ten independent components each have probability 0.2 of failing a specified test, and X counts failures. Under those assumptions, X follows a binomial model with n = 10 and p = 0.2. The probability of exactly two failures is (10 choose 2)(0.2)2(0.8)8 ≈ 0.3020.

The probability of at least one failure is easier through the complement: 1 − P(X = 0) = 1 − (0.8)10 ≈ 0.8926. The expected number is np = 2, and the variance is np(1 − p) = 1.6. An expected value of two does not mean every batch contains exactly two failures.

The assumptions deserve scrutiny. If components share a manufacturing fault, their failures may be related. If the probability differs sharply between components, one common p may be inappropriate. Naming the binomial distribution is not enough; the question’s situation must support its structure.

Worked example: expectation need not be an attainable outcome

An invented game awards X points, where X takes values 0, 2 and 5 with probabilities 0.5, 0.3 and 0.2. Then E(X) = 0(0.5) + 2(0.3) + 5(0.2) = 1.6. Also E(X2) = 0 + 4(0.3) + 25(0.2) = 6.2, so Var(X) = 6.2 − 1.62 = 3.64.

No single play awards 1.6 points. The expectation is a weighted average across the model. If a bonus of three points is added to every result, the mean becomes 4.6 while the variance stays 3.64. A constant shift changes location but not spread. If all scores are doubled, the variance is multiplied by four.

For calculator practice, write the event first: P(X ≥ 3), for example. Then translate it into the calculator’s cumulative convention if necessary. Finish by checking whether the complement boundary is 2 or 3. Off-by-one errors are reasoning errors with a very small visual footprint.

Contents · Previous chapter · Next chapter

CHAPTER 13 OF 16

Distinguish an individual measurement from a sample mean

Back to contents

A normal model describes a continuous quantity using a mean and a spread. A sample mean is a different random variable: it averages several observations. Confusing the two produces answers that can look plausible while using the wrong uncertainty.

Worked example: one package

Suppose a package mass X is modelled as normal with mean 500 g and standard deviation 4 g. Write X ~ N(500, 16), because the second parameter in this notation is the variance. For P(X < 496), standardising gives z = (496 − 500)/4 = −1. The probability is approximately 0.1587.

A sketch helps: 496 is one standard deviation below the mean, so the shaded left tail should be smaller than half the distribution. If a calculator returns 0.8413, the likely problem is the selected tail rather than the normal model itself.

Worked example: an average of packages

Now take a random sample of 16 independent packages from that same normal population. The sample mean has mean 500 g and standard deviation 4/√16 = 1 g. Thus P(the sample mean < 496) corresponds to z = −4, approximately 0.0000317. A single package below 496 g is much more likely than an average of 16 packages below that value.

Averaging reduces random fluctuation because unusually high and low observations can offset each other. It does not remove bias. If the weighing instrument is miscalibrated, taking more measurements with that instrument may produce a very precise average of the wrong quantity.

When the population itself is not normal

The central limit theorem provides a route to an approximate normal model for an average of sufficiently many independent observations from the same population, when its variance is finite. Individual waiting times might be strongly skewed, while averages across large random samples have a much more nearly bell-shaped distribution. This does not make the original waiting times normal. The object being approximated is the sample mean.

Suppose such a population has mean 20 minutes and standard deviation 8 minutes. For a sample of 64 independent observations, the mean still centres on 20, while its standard deviation is 8/√64 = 1 minute. A normal calculation for that mean is an approximation whose suitability depends on the sampling conditions and population shape. More extreme skewness or outliers can require more caution. State the justification supplied by the question, rather than treating sample size as a magic switch.

Sampling is a design question

Imagine asking only the first twenty students entering a library how many hours they study. The arithmetic mean can be calculated perfectly, but the selection method may favour students with particular habits. A large convenient sample is not automatically representative of the population of interest.

When working with a sample, state the population, the selection process and the variable measured. Then distinguish the observed sample statistics from the unknown population quantities they estimate. The sample mean is a number once data have been collected; before sampling, it is a random variable that could take different values.

Check the variance calculation

For observations 8, 10 and 12, the sample mean is 10. The squared deviations sum to 8. The unbiased estimate of population variance is 8/(3 − 1) = 4. Dividing by 3 would give the mean squared deviation of this particular data set, a different quantity. Label the target before choosing the denominator.

The practical study habit is simple: write “individual”, “sum” or “mean” beside the random variable before entering any parameters. That one-word label helps protect the square root, the variance and the interpretation throughout the solution.

Contents · Previous chapter · Next chapter

CHAPTER 14 OF 16

Use hypothesis tests to evaluate evidence carefully

Back to contents

A hypothesis test asks how surprising the observed result would be under a stated baseline model. It does not calculate the probability that the baseline claim is true. This distinction prevents overconfident conclusions and makes the final sentence of a test as important as its numerical calculation.

Worked example: has a mean increased?

Suppose a normally distributed measurement has known population standard deviation 12 units. A random sample of 36 independent observations has mean 104. We wish to test whether the population mean exceeds 100. Set H0: μ = 100 and H1: μ > 100.

Under H0, the sample mean has standard deviation 12/√36 = 2. The observed test statistic is z = (104 − 100)/2 = 2. The upper-tail probability is approximately 0.0228. At a 5% significance level, this is small enough to reject H0. There is sufficient evidence, under the model and sampling assumptions, that the population mean exceeds 100.

The conclusion does not say that 97.72% of measurements exceed 100, or that there is a 97.72% probability the alternative is true. The p-value concerns the extremeness of the sample result under the null model. Keep the population claim, sample observation and probability statement separate.

Choose the direction from the question

If the question asked whether the population mean had changed in either direction, the alternative would be μ ≠ 100. For this symmetric normal test, the two-tailed p-value would be approximately 0.0455. The direction must be selected from the investigation’s question, not chosen afterwards to obtain a more favourable result.

A failure to reject the null is not proof that the means are exactly equal. The evidence may be insufficient to distinguish the proposed change from sampling variation. A careful conclusion says what the test supports, using the chosen significance level, and avoids claiming more.

A useful answer structure

Define μ in context, state the hypotheses, identify the distribution used under the null, calculate the relevant probability or critical comparison, state the decision and write the contextual conclusion. This order makes the argument readable. A correct calculator value without a clear population claim leaves the reader unsure what has been tested.

For revision, change one feature of the example at a time: sample size, observed mean, significance level or the direction of the alternative. Predict the effect before calculating. This develops judgement about evidence instead of turning testing into a sequence of memorised buttons.

Contents · Previous chapter · Next chapter

CHAPTER 15 OF 16

Use regression for a defined prediction task

Back to contents

Regression describes a relationship in observed data and can support a prediction within a suitable range. It does not explain a cause by itself. Before calculating anything, identify which variable is being used to predict which other variable and inspect the scatter of points.

Worked example: interpret a fitted line

Suppose an invented data set relating practice time x in hours to task completion time y in minutes produces the fitted line y = 30 − 2x. For x = 4, the predicted completion time is 22 minutes. The slope means the fitted model associates one extra practice hour with a two-minute reduction in predicted completion time.

If the data were collected for practice times between one and six hours, predicting at four hours is interpolation. Predicting at twenty hours is extrapolation and would produce a nonsensical negative completion time. The equation’s algebraic availability is broader than its credible use.

The intercept of 30 is the model’s prediction at zero practice hours. Whether that interpretation is useful depends on the data range and the context. A coefficient can have a formal meaning without being well supported by observations at that value.

Direction matters

A line fitted to predict y from x minimises vertical discrepancies in y. It should not automatically be rearranged to predict x from y. Reversing the prediction task generally requires the appropriate regression of x on y. A student should write the target variable before selecting the calculator command.

Correlation summarises the strength and direction of a linear association. A value near zero does not rule out a strong curved relationship. A large positive or negative value does not establish that changing one variable causes the other to change. Common influences, selection effects or coincidence may still matter.

Transformations can make a pattern easier to model

If a positive response appears to grow exponentially, plotting ln y against x may reveal an approximately straight pattern. A fitted relation ln y = a + bx corresponds to y = eaebx. The transformation changes how the model is expressed, so the final prediction must be converted back to the original units.

Good statistical writing names the relationship, the range in which it was observed and the limitations of the prediction. Those sentences make a numerical answer useful to someone making a decision. They also help a student notice when a calculation has answered a different question from the one asked.

Contents · Previous chapter · Next chapter

CHAPTER 16 OF 16

Build a study plan that produces independent solutions

Back to contents

A useful plan has a learning task, a checking task and a return visit. The learning task might repair inverse functions. The checking task might be an unseen question with a changed domain. The return visit might happen several days later without notes. Completing all three provides better evidence than immediately repeating an example while its solution is still visible.

A practical weekly cycle

SessionTaskEvidence to keep
UnderstandStudy one idea and explain a worked example.A short explanation of why the method fits.
PractiseSolve related questions with changed numbers and conditions.Working completed before consulting solutions.
MixAttempt questions from several previously learned topics.A record of method-selection difficulties.
ReturnRedo one earlier error from a blank page.A corrected solution and a specific checking habit.
An adaptable practice cycle, not a school requirement

Scale the amount to the student’s timetable. A short session with a clear purpose can be more informative than an exhausted evening of copying. During a busy school week, maintain contact with older topics through a few well-chosen questions. During a quieter period, use longer sessions for mixed problems and timed sections.

Read a marked script as evidence

Choose three lost-mark examples and classify them precisely. Was the domain ignored? Was the wrong distribution selected? Was the derivative correct but the stationary point misinterpreted? Then choose one repair for each category. Avoid a large undifferentiated pile of “redo” questions, which can conceal whether the same problem is recurring.

Parents can ask, “Which step will you check differently next time?” This keeps the conversation on an action the student can control. Compare independent work over time: the same type of question, similar conditions and a clear record of help received. A heavily guided correct answer and an independent correct answer represent different stages of learning.

Frequently asked questions

Does a weak first test mean H2 Mathematics is unsuitable? A single result cannot settle that question. Examine the errors, the student’s response to feedback, workload and school advice. Some difficulties concern specific foundations that can be repaired; others may reveal a mismatch between current preparation and available support. The decision should use a pattern of evidence.

Should students finish every question in a large question bank? Volume alone does not show learning. Choose enough questions to encounter meaningful variations, then include mixed and delayed attempts. If several near-identical questions are already secure, use the time to address a different weakness.

Can a graphing calculator replace written reasoning? It can support calculation and investigation, but the question determines what must be shown. Practise communicating the model, essential steps and interpretation. A numerical output becomes useful when the reader can see what it represents.

Should JC1 students begin full papers immediately? Full papers contain material that may not yet have been taught. Use the school’s topic coverage to select appropriate questions. Build towards whole-paper practice as coverage expands, while retaining targeted work on persistent gaps.

How can students tell whether they understand a solution? Close it, explain why the first method was chosen, and solve a related problem with one condition changed. If the method only works when the original layout remains visible, more independent practice is needed.

Where should families check official details? Use the SEAB 2027 school-candidate syllabus list, the linked subject document and the school’s examination guidance. This guide supplies original explanations and practice examples; the official documents determine the examination requirements.

The next useful step is small: select one chapter that matches a current difficulty, attempt its example without looking, and explain the answer aloud. From there, change one condition and try again. H2 Mathematics becomes more manageable when a student can see both the structure of a problem and the next move that structure permits.

Contents · Previous chapter